Why you need an AI Stoplight

Luke Allpress

September 02, 2023

Like every other school district in the country world, we  were taken by surprise by the arrival of ChatGPT and other AI chatbot competitors in late 2022. We reflexively blocked the OpenAI tools on our network and on student devices, sent a rushed email out to all staff that "we are aware of the power of these tools, please know that blah blah blah" and tried not to think about it for a few months.

Once we realized that the ostrich approach would not be enough for our staff or our students, we began to tap our networks and search the internet (and yes, ask ChatGPT) for advice and best practices. What policies do we need to have in place? What are we getting wrong about this? What are the opportunities we are missing?

In April 2023, there weren't a lot of answers to these questions to be found. We called a meeting of the best folks we know (current teachers) and listened while they talked it out in groups by topic:

1. 1. 1. How are students using these tools, and what does that mean for lessons? The good and the bad.

      2. What skills are needed for our students to succeed in a world of Generative AI?

      3. #teacherhacks, how can this improve your job.

We took notes on everything we heard, created a group chat space to continue the conversation, and went off to workshop it.

Besides crafting a backstop policy for teachers (our own version of "any use of GAI without teacher permission constitutes academic dishonesty) we still had more questions than answers. 

Students might use these tools to "cheat" on assignments we want them to do using their own knowledge. Okay, is the problem the tool or the assignment?

Students need critical thinking, collaboration, communication, and prompt engineering skills in order to succeed. Hmm, except for prompt engineering, that sounds like the "21st century skills" I've heard a lot about for...decades?

The most interesting finding though was that when we had 35+ folks in the room who chose to be there, who opted-in, the conversation took off. When we tried to have follow-up conversations with teachers who hadn't opted in, whose only exposure to GAI was "that one 60 Minutes episode", we were two ships passing in the night. There was nothing to talk about, no common language.

After flagging this need, and honing in on one specific line we heard from a teacher ("It will be incredibly important to distinguish what constitutes cheating and what citing this tool will look like in the classroom") we began talking through how we could accelerate join-up, get the uninitiated to the point where we could have a conversation about the risks and rewards GAI brings to the table.

Our of hours of talk came the AI Stoplight. Kicked off by this image from DitchThatTextbook, we figured out that not only do we need to talk about "what is cheating" anymore, we need to build in capacity for that to change by teachers as needed for each assignment. We needed categories, a gradient, from "any AI is cheating" to "all AI is welcome" and after talking through 2 to 5 categories, we landed on 3, and the Stoplight analogy wrote itself from there.

"It will be incredibly important to distinguish what constitutes cheating and what citing this tool will look like in the classroom"

After flagging this need and honing in on one specific line we heard from a teacher (above) we began talking through how we could accelerate join-up, get the uninitiated to the point where we could have a conversation about the risks and rewards GAI brings to the table.

Out of hours of talk came the AI Stoplight. Kicked off by this image from DitchThatTextbook, we figured out that not only do we need to talk about "what is cheating" in this world, we need to build in capacity for that definition to change as needed for each assignment. We needed categories, a gradient, from "any AI is cheating" to "all AI is welcome" and after talking through 2 to 5 categories, we landed on 3, and the Stoplight analogy wrote itself from there.

This simple framework has allowed us to anchor conversations with teachers and district leaders in simple, comprehensible language. We have rolled out slides to teachers to run lessons for all 10k students in our district, teaching them about the stoplight levels and what different lessons/activities look like in each color.

Implicit in this stoplight is that we now have opened up full access to GAI tools for our students. We did not come to this decision lightly, and are being thoughtful and intentional with student data and ethical concerns, but we ultimately made the call that these skills will be necessary for students' futures, and we need to find ways to teach them safely and effectively.

Do we have great lesson examples for each bot yet? Absolutely not, especially green bot. We're always eager to collect and share out great ideas, so if you try or see any great examples please share! Resources can be shared to aiforlearningcontact@gmail.com or tagged on socials with #aistoplight, to be shared in future newsletters.